Purpose This study investigates the preparedness of archives and records management (ARM) professionals in Nigeria for the adoption of artificial intelligence (AI) technologies. It also examines the influence of institutional drivers on AI integration efforts. The research addresses a critical gap in localized understanding of how individual and organizational readiness intersect to shape AI adoption in the Nigerian records management sector. Design/methodology/approach The study employed a cross-sectional survey design targeting ARM professionals registered with the Society of Nigerian Archivists. A structured questionnaire was administered via an online platform. Data from 217 respondents were analysed using descriptive statistics, t-tests and Pearson correlation to assess levels of awareness, preparedness, institutional support and perceived barriers. Findings The findings reveal individual preparedness, positive perceptions of AI benefits and strong willingness to adopt AI if trained. However, institutional support was weak, particularly in leadership engagement, strategic planning and funding. A significant positive correlation was found between institutional drivers and AI preparedness, while no gender-based differences in readiness were observed. Research limitations/implications The study emphasizes the importance of institutional reform in driving AI readiness. It suggests that national and organizational policies should prioritize AI training, leadership commitment and infrastructural investment. The results offer a basis for evidence-driven policy, professional development and curriculum design to support digital transformation in records management. Originality/value This research is among the first empirical studies to assess AI adoption readiness specifically within Nigeria’s records management sector. It contributes a localized understanding of how individual competencies and institutional enablers jointly affect AI integration. The findings provide a foundation for comparative research across similar contexts in developing nations.
Purpose Artificial intelligence (AI) can transform library services by enhancing efficiency and user experience. The adoption of AI in Pakistani libraries remains underexplored. Design/methodology/approach This study aims to identify the determinants of AI adoption intention among library professionals using an extended unified theory of acceptance and use of technology (UTAUT) model integrated with combined importance-performance map analysis (cIPMA). Findings The key constructs, such as effort expectancy, performance expectancy, relative advantage, top management support, trust and competitive pressure, were assessed using partial least squares structural equation modeling. The findings highlight effort expectancy and relative advantage as significant predictors. Originality/value The study contributes to theoretical advancement by extending UTAUT in a library context and offers practical insights to help administrators facilitate AI adoption.
Purpose This study aims to comparatively evaluate the retrieval performance of select web search engines, i.e. Google, Bing and Yandex, based on standardised metrics. Design/methodology/approach The study draws informational queries from the discipline of business studies using titles, keywords and abstracts of PhD theses from the Shodhganga repository. It categorises queries into one-word, two-word and complex-word queries and evaluates them using pre-established relevance criteria on a three-point relevance scale to calculate precision, mean average precision, relative recall, F-measure and fall-out ratios. The study further employs ANOVA to determine the statistical significance of the difference in means between the above-stated metrics for Google, Bing and Yandex, to understand the performance of one web search engine compared to the others. Findings The analysis of variance (ANOVA) results indicate that the difference between the means is not statistically significant for Google, Bing and Yandex in terms of precision, relative recall, F-measure and fall-out ratio. However, descriptive statistics, although not statistically significant, revealed a superiority of Google over the other two search engines in terms of mean average precision, mean relative recall, F-measure and fall-out ratio. Additionally, it was found that complex-word queries performed better than one-word and two-word queries for all three web search engines. Research limitations/implications The paper studies the retrieval effectiveness of only three web search engines, among a pool of search engines. It restricts its use to only informational queries, leaving aside the navigational and transactional queries. Future studies can target a broader range of web search engines, utilising all three categories of queries, including navigational and transactional queries, across multiple disciplines. The study has implications for users, web search engine programmers and Librarians. It will help users, especially academic users in the field of business studies, to identify more suitable web search based on the results of the metrics, as well as help programmers to understand which retrieval techniques work better and thereby improve the efficiency of search engines better. Moreover, librarians include topics such as retrieval effectiveness in web search, query formulation strategies, keyword mappings, etc., in their library orientation programs. Originality/value The study provides an original empirical evaluation of the performance of web search engines Google, Bing and Yandex by conducting a statistical comparison of their performance on informational queries drawn from the field of business studies. It also compares the select web search engines based on query types.
Purpose The study aimed to identify the effects of emerging technologies on sustainable library services. In this study, sustainability refers to the long-term efficiency, adaptability, accessibility and user-centeredness of library services enabled through technological integration. Design/methodology/approach This study used a quantitative research design to address the objectives. The population consisted of librarians from the public and private sector universities of Pakistan. The data were analyzed using SmartPLS software. Findings The analysis consisted of two major parts: first, the assessment of measurement model and second, the structural equation modelling analysis. A significant positive effect of emerging technologies (artificial intelligence, gamification, wearable technologies, Internet of Things, blockchain technology, ChatGPT and m-learning) was found on the sustainable library services. Research limitations/implications The study has provided worthwhile theoretical contributions by adding valuable literature to the existing body of knowledge. Practical implications It has provided practical contributions by offering a context-specific baseline for management bodies in university libraries, particularly in developing countries, to construct policies for the successful adoption of emerging technologies for sustainable library services. Social implications The study has social implications too as smart library services prove fruitful for society. Originality/value This is the first empirical study examining the integrated effects of multiple emerging technologies on sustainable library services within the context of university libraries in Pakistan. Its implications may be cautiously extended to similar developing academic environments.
Purpose Identifying similarities between parts of a text, paper or document has implications for several information processing and retrieval tasks. Thus, information processing and retrieval research aims to detect content relations in terms of lexical, conceptual, semantic and content features. Semantic similarity measures refer to detecting content relations based on the equivalence between texts using the semantic component. Measuring text similarity is the aim of our work.Design/methodology/approach This work involves the following main steps: Thorough data preprocessing is crucial. This involves removing stop words, common words that add little semantic value. Tokenization is the process by which a large amount of text is divided into smaller parts (phrases) called tokens. Applying stemming or lemmatization also helps reduce words to their base form, thereby standardizing the dataset for better comparison. The various attribute types are used. Phrases are the attributes that characterize the concepts used in the text. Cosine indexes generate binary values between 0 and 1. However, cosine reflects the similarity in terms of percentage. When cosine similarity is near 100, it refers to the binary value 1. The cosine similarity focuses on the angle between two vectors. This makes cosine similarity more robust in capturing the pattern similarities between two data sets, even if their magnitudes differ. (Phil Miesle 2023). The phrase length is measured by the number of words in a phrase without stop words. The phrase length normally ranges from two to eight. If the phrase length is longer, the phrase is content expressive and far from generic. The phrase length is the expression of concept and semantic richness. The literature does not address the phrase length and its impact on text similarity and other features. The available research mainly outlines the linguistic attributes; for example, the lexically derived syntactic representations are discussed (Stallings, 1998) and the structure of surrounding phrase lengths (Zvonik, 2003). Removing stop words from sentences reduces the low-level information, which makes the sentences easy to understand. Moreover, removing the stop words doesn't affect the contextual meaning of sentences, as machines preserve the semantic information of words. (Han, 2021; Hasanah, 2019; Hu, 2022; Huang, 2019) Finally, the similarity is calculated as the minimum distance path between nodes (phrases), which are cosine. We evaluated the phrase similarity using relatively small benchmark datasets. A significant scientific phrase, "Metabolic Syndrome", is used to search Google Scholar and a few highly cited papers are identified. Each selected full text is subjected to phrase extraction, ranking phrases based on frequency. These phrases contain stop words that have been removed. The phrases contain a minimum of two words and a maximum of eight words. When stop words are removed, the content of the treated concept is preserved, and when frequent phrases are selected, they are considered tokens that represent the content. The top frequent phrases are selected for each sample, and the similarity is measured using cosine.Findings Even though the lexical and semantic similarity measures are different, lexical and semantic similarities are correlated in highly specialized documents such as journals or scientific reports. Cosine similarity measures report a low-level correlation even if there is no common word between two phrases. For example, the phrases "glance disease models and mechanisms" and "insulin resistance visceral adiposity atherogenic dyslipidemia and endothelial" have no common words and still exhibit a cosine similarity score of 24.4%. Partial phrase similarity is observed in different sets of data in this work. The analysis reveals that in the highly frequent phrases, the most commonly used words are insulin, syndrome, cardiovascular, diabetics, and therapeutics, followed by a few next-level supporting words such as resistance, models, mechanisms, treatment and systems. Thus, we can understand how a core theme is treated in the literature in supporting terms.Research limitations/implications This study has a few limitations. First, the dataset is relatively small and focuses exclusively on highly cited papers within a single medical domain, which may limit its generalizability to the broader scientific literature. Second, the analysis filters out low-frequency phrases, concentrating solely on high-frequency terms. This exclusion may overlook emerging concepts that appear infrequently but hold significant semantic value. Third, while cosine similarity effectively captures structural relationships, it may not fully account for deeper cognitive or contextual connections, particularly when phrases share no common words yet remain conceptually linked. Additionally, the impact of phrase length on similarity measures remains underexplored, limiting the framework's ability to systematically assess how syntactic complexity influences scoring. Finally, reliance on static phrase extraction may oversimplify dynamic linguistic variations, highlighting the need for more adaptive NLP techniques in future work.Practical implications The dataset used is smaller in terms of phrases analyzed. The low-frequent phrases are removed from the analysis, leading to the processing of selected high-frequent words. However, this work indicates the use of phrases for processing. Cosine values are more productive when measuring semantic similarity than lexical similarity. Our future work will determine the relation between phrase frequency and cosine similarity values to generate strong inferences. We intend to extend the work with more samples and datasets and include more phrases, which can lead to effective results.Social implications Detecting phrase relations, both complete and partial, will enable the identification of document similarities. The existing practices are often unrelated to finding cognitive and semantic relations. This kind of exercise leads us to pave the way for measuring naturally existing similarities among literature and sub-domains.Originality/value This paper identifies the naturally existing relations between concepts in natural language texts. Text analysis will be improved using the conceptual model described in this paper.
